Cathepsin D, a Lysosomal Protease, Regulates ABCA1-mediated Lipid Efflux*

To identify genes involved in the regulation of plasma high density lipoprotein (HDL) cholesterol (HDL-C) levels, patients with low HDL-C and age- and sex-matched controls (normal HDL-C) were extensively characterized. Comparative transcriptome analysis was carried out in cholesterol-loaded monocyte-derived macrophages from low HDL subjects segregated into groups with or without cholesterol efflux defects or ABCA1 mutations. Clusters of differentially regulated genes were evident in the low HDL groups as compared with controls. Of particular note, expression of cathepsin D (CTSD), a lysosomal proteinase, was reduced by ∼50% in monocyte-derived macrophages of low HDL-C subjects, most significantly those with cholesterol efflux defects but without mutations in ABCA1 (p < 0.01). These results were verified by reverse transcription-PCR and replicated in a second cohort. We show here that blocking the activity or expression of CTSD, by pepstatin or CTSD small interfering RNA, respectively, reduced ABCA1 expression and protein abundance in both macrophages and CHO cells and apolipoprotein A-I-mediated lipid efflux by more than 70%. Conversely, expression of CTSD increased both ABCA1 mRNA expression and cellular ABCA1 protein. Consistent with its role in the proteolytic processing of prosaposin, inactivation of CTSD function resulted in the accumulation of glycosphingo-lipid and free cholesterol in late endosomes/lysosomes, a phenotype similar to NPC1 deficiency. Inhibition of CTSD also caused retention of ABCA1 in lysosomal compartments, reducing its trafficking to the plasma membrane. These studies demonstrate a novel and potentially important role for CTSD in intracellular cholesterol trafficking and ABCA1-mediated efflux. Therefore, decreased CTSD expression may contribute to low plasma HDL-C levels.

High density lipoprotein cholesterol (HDL-C) 3 deficiency is a major risk factor for cardiovascular diseases and has a complex metabolic and genetic etiology. Overproduction and/or impaired clearance of triglyceride-rich lipoproteins results in triglyceride enrichment of HDL and accelerated apoA-I catabolism and may account for low HDL-C in patients with other features of the metabolic syndrome. Low HDL can also result from mutations in a number of genes regulating HDL production or catabolism, including apolipoprotein A-I (1-4), lecithin:cholesterol acyltransferase (5-7), phospholipid transfer protein (8 -10), and ATP-binding cassette protein A1 (11)(12)(13)(14)(15)(16)(17)(18)(19).
Defective cellular phospholipid and cholesterol efflux is also a common contributory phenotype in familial low HDL syndromes, and in several of these kindreds, affected subjects are heterozygous carriers of ABCA1 mutations (17,20,21). These genetic studies demonstrate the key role of ABCA1 in cholesterol efflux to apoA-I, but as we have reported previously (22), sequence variants in ABCA1 account for a small proportion of hypoalphalipoproteinemia. In further studies in a cohort of low HDL and control subjects previously characterized (Ref. 22 and Sarov-Blat et al. 4 and Kiss et al., 5 submitted manuscripts), we have undertaken to identify other genes and gene products that contribute to lipid efflux and low HDL. High density oligonucleotide microarrays were used to analyze gene expression in cholesterol-loaded monocyte-derived macrophages (MDM) prepared from these subjects. Importantly, we demonstrate here that expression of cathepsin D (CTSD), a lysosomal proteinase (23), is reduced by 50% in the MDM of low HDL-C subjects, significantly in those with cholesterol efflux defects without mutations in ABCA1, compared with control subjects. CTSD is involved in intracellular catabolism in lysosomal compartments but also functions in hormone and antigen processing (24 -26). CTSD has been implicated in mitogenic activity in certain cancers (27,28), which may be related to the secretion of its pro-form (24), lysosomal membrane permeabilization, and apoptosis (29). In addition, genetic variants in CTSD have been identified as a risk factor for Alzheimer disease (30) and are the causative gene of neuronal ceroid lipofuscinosis (31,32). CTSD has also been previously shown to bind to and promote the maturation of prosaposin (pro-SAP) into saposins (A, B, C, and D), which are required for the hydrolysis of certain sphingolipids (33,34). We further demonstrate that CTSD thereby regulates intracellular sphingolipid metabolism, cholesterol trafficking, and ABCA1-mediated efflux. These findings are relevant to our understanding of the cellular etiology of certain low HDL states, with specific reference to the regulation of lysosomal enzymes involved in the intracellular cholesterol trafficking and therefore apoA-I/ ABCA1-mediated cellular cholesterol efflux.
Selection of Low HDL-C Subjects and Controls-Patients referred to the Lipid Clinic of the University of Ottawa Heart Institute were selected based on the following criteria: Caucasian, HDL-C Ͻ5th percentile for age and sex; triglyceride Ͻ95th percentile (most Ͻ75th percentile); and LDL cholesterol Ͻ75th percentile, measured while not on treatment with lipidmodifying medications. Exclusion criteria included clinical conditions or medications causative of low HDL-C, including all forms of diabetes. There was no other selection bias, since most subjects were recruited in chronologic order of referral. The control group was selected among healthy and normolipemic volunteers of the same ethnic background recruited from the Ottawa region. The study was approved by the University of Ottawa Heart Institute Human Research Ethics Committee, and written informed consent was obtained from all participants.
RNA Isolation in Human Subjects-Macrophages were homogenized with Qiashredder, and total mRNA was isolated using the RNeasy mRNA isolation kit (according to the manufacturer's instructions).
GeneChip Affymetrix Microarray-Gene expression profiles were analyzed by Affymetrix GeneChips, which cover 22,215 annotated genes. RNA quality and quantity was assessed with an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA) and RiboGreen RNA quantization reagent (Molecular Probes, Inc., Eugene, OR), respectively, in addition to measuring OD at 260 and 280 nm and 260/280 ratios. All samples possessed 18 and 28 S rRNA bands with no sign of RNA degradation. Affymetrix human U133A GeneChips (Affymetrix Inc., Santa Clara, CA) were used. Briefly, double-stranded cDNA was synthesized from each total RNA sample via oligo-T7-mediated reverse transcription. Biotin-labeled cRNA was synthesized using an in vitro transcription reaction with the Enzo BioArray high yield RNA transcript labeling kit (Enzo Diagnostics, Inc., Farmingdale, NY). Subsequently, the purified, biotin-labeled cRNA was fragmented into 50 -150-nucleotide fragments. Fragmented, biotinlabeled cRNA (15 mg), together with a known spike (according to Affymetrix recommendations) was hybridized to a U133A Gene-Chip for 16 -18 h at 45°C. An Affymetrix GeneChip fluidics station 450 was used to perform standard washing and streptavidin/ phycoerythrin staining. The chips were scanned using an Affymetrix GeneChip scanner 3000.
TaqMan Quantitative Real Time PCR-Total RNA (1 g) was converted to cDNA using a high capacity cDNA archive kit (Applied Biosystems), and the equivalent of 10 ng/well was arrayed into high density 384-well plates using a Biomek FX robot (Beckman Coulter), allowing expression profile generation for one gene at a time. Quantitative RT-PCR was carried out using a 7900HT Sequence Detector System (Applied Biosystems) in a 10-l reaction volume. TaqMan Universal PCR FIGURE 1. Cluster of the cathepsin family genes whose transcripts were decreased in low HDL groups (group A, low HDL, low efflux, no ABCA1 mutation; group B, low HDL, low efflux, ABCA1 mutation; group C, low HDL, normal efflux compared with control group D). Monocyte-derived macrophages isolated from patients with low HDL and age-and sex-matched controls were loaded with Ac-LDL, and global gene expression profiles were analyzed on duplicate GeneChip DNA-microarrays. A, hierarchical cluster dendrogram from TIGR MeV program for both genes and subjects. The cathepsin gene family was expressed similarly within low HDL groups and between control subjects. The red color represents up-regulated genes, whereas the green indicates down-regulated genes. The CTSD gene was the most differentially expressed member of the cathepsin genes and was significantly down-regulated in groups A and B compared with control (p Ͻ 0.01 and p Ͻ 0.016, respectively). B, plot of the CTSD probe set relative expression in microarray for individual subjects and average CTSD mRNA expression value for each group.
Master Mix 2X (Applied Biosystems) and universal PCR conditions recommended by the vendor were followed.
SYBR Green Real Time PCR-mRNA was isolated from mammalian cells using Trizol reagent (Invitrogen, Burlington, Canada). CTSD mRNA, ABCA1, and GAPDH mRNA were quantified using the Stratagene SYBR Green kit on a Lightcycler real time PCR machine according to the manufacturer's recommendations.
Cell Culture-Human blood monocytes were differentiated into macrophagic cells according to Cohen et al. (22). The monocytes were differentiated into macrophages over the course of 12 days in growth media, and the resulting cells were used in the efflux assay or for mRNA isolation. All macrophages were grown in RPMI medium supplemented with 10% pooled human serum, whereas J774 and CHO cells were grown in Dulbecco's modified Eagle's medium (DMEM) with 10% fetal bovine serum (FBS). All cells were grown at 37°C in 5% CO 2 and incubated with penicillin/streptomycin. Bone marrow-derived macrophages were flushed from mouse femurs. Macrophages were generated by incubating fetal liver cells (2 ϫ 10 6 cells/ml) or bone marrow cells (10 6 cells/ml) with DMEM, 10% FBS complemented with 15% L929 conditioned medium for 7 days. Three days prior to the isolation of peritoneal macrophages, mice were injected with 2 ml of 3% thioglycollate intraperitoneally. The animals were then euthanized with CO 2 , the peritoneal cavity was washed by an 8-ml injection of saline, and the cells were washed three times by centrifugation at 1000 ϫ g for 10 min with DMEM. The cells are then plated in 6-well plates containing DMEM, 10% FBS overnight.

H]Cholesterol and [ 3 H]Phospholipid Efflux-Macrophages and
CHO cells were labeled with LDL or Ac-LDL (50 g/ml) equilibrated with [ 3 H]cholesterol (5 Ci/ml) in 1% FBS medium for 24 h. The next day, the cells were washed twice with phosphate-buffered saline (PBS) and equilibrated overnight in serum-free medium. ApoA-I cholesterol efflux was monitored by counting the radioactivity in the medium after a 24-h incubation of 10 g/ml lipid-free ApoA-I, 0.1% BSA. Cholesterol efflux was expressed as the percentage of [ 3 H]cholesterol in the medium over total cellular [ 3 H]cholesterol. For phospholipid efflux, the same protocol was used as for cholesterol efflux, except that the 3 H-labeled phospholipids bound to immunoprecipitated apoA-I were counted, using a rabbit polyclonal anti-human apoA-I antibody. Since some CTSD is secreted by cells and could degrade protein in the medium, we analyzed by SDS-PAGE immunoprecipitated apoA-I from the medium after the efflux assay. No apoA-I degradation was observed (data not shown). Lipid Analysis-Free cholesterol and cholesteryl esters were extracted and separated by TLC on Sil-G TLC plates (EMD Chemicals, Darmstadt, Germany) using the solvent system hexane/diethylether/acetic acid (105:45:1.5) (35). Lipid bands were visualized with iodine vapors, isolated, and analyzed by ␤-scintillation counting.
siRNA Transfection and Expression Vectors-Cathepsin D siRNA obtained from Ambion (Austin, TX) was transfected into macrophage J774 cells using the TransIT-TKO transfection reagent (Mirus, Madison, WI). Briefly, 100 nM siRNA and 4 l of the transfection reagent were both mixed with 100 l of serum-free DMEM for 5 min. The solutions were then incubated at room temperature for 20 min. This mixture was applied to the DMEM, 10% FBS medium in each well of a 12-well plate (for cholesterol efflux) or 6-well plate (for RT-PCR and Western blotting) and incubated for 36 h. J774 or CHO cells were transiently transfected with human cathepsin D vector using Lipofectamine 2000 according to the manufacturer's recommendations.
Cell Surface Biotinylation and Western Blotting-Biotinylation was performed as described earlier (36) with slight modification. J774 or CHO cells were treated with or without pepstatin for 36 h, and surface proteins were biotinylated with 500 g/ml sulfosuccinimidyl 2-(biotinamido)-ethyl-1,3-dithiopropionate (Pierce) for 30 min at 4°C on a platform rotator. Cells were then washed with ice-cold quench buffer (1 M Tris-HCl (pH 7.5)) and once with ice-cold phosphate-buffered saline. The cells were swollen on ice for 10 min and then homogenized with 20 strokes in a tight fitting Dounce homogenizer. After centrifugation at 2000 ϫ g, at 4°C, for 10 min to remove unbroken cells and nuclei, the supernatant was recentrifuged at 100,000 ϫ g, at 4°C, for 60 min. The resulting supernatant was discarded, and the final membrane pellet was resuspended in 250 l of immunoprecipitation buffer; 150 g of protein was added to 30 l of streptavidin-Sepharose (Amersham Biosciences) and incubated overnight on a platform mixer at 4°C. The gel was pelleted and washed five times with immunoprecipitation buffer. The Western blotting technique was performed as described before (36). Briefly, cells were washed twice with ice-cold PBS and scraped into 0.5 ml of radioimmune precipitation buffer and homogenized with 20 strokes in a tight fitting Dounce homogenizer. After centrifugation at 2000 ϫ g, at 4°C, for 10 min to remove unbroken cells and nuclei, the protein concentration was determined by a Lowry assay. The same amounts of proteins were electrophoresed on a SDS-polyacrylamide gel and transferred to nitrocellulose at 125 V for 4 h. The nitrocellulose membranes were then blocked with 5% milk in Tris-base-buffered solution and incubated overnight with CTSD, HSP60, ABCA1, or prosaposin antibodies. Horseradish peroxidase-linked secondary antibodies (Amersham Biosciences) were used, and bands were detected by chemiluminescence (Supersignal substrate; Pierce).
Fluorescence Microscopy-Cells were seeded on 35-mm dishes with attached glass coverslips (Fisher) and transfected with ABCAI-GFP. The coverslips were then mounted in a temperature-controlled chamber (37°C) in regular growth medium supplemented with 20 mM HEPES (pH 7.4) and visualized with an Olympus ϫ100 oil immersion objective, numerical aperture 1.4, on an Olympus IX80 laser-scanning confocal microscope operated by FV1000 software version 1.4a. The GFP was excited with a 515-nm laser, and the BODIPY-lactosylceramide was excited with the 488-nm line of a multiple-line argon laser, and the Lysotracker red (DND-99; Molecular Probes) was excited with the 543-nm line of a helium/neon green laser. Alternatively, for analysis of subcellular distribution of cholesterol, cells were seeded on glass coverslips similar to those above and incubated with regular growth medium with 10% FBS. 24 h later, cells were washed two times quickly with PBS and incubated with 200 l of 3.3% paraformaldehyde for 10 min. The paraformaldehyde was removed, and the cells were rinsed once with PBS and then quenched for 30 min with 50 mM NH 4 Cl. The coverslips were then visualized with an Olympus ϫ100 oil immersion objective, numerical aperture 1.35, on an Olympus 1X70 inverted microscope operated by TILLvisION software version 4.0 (TILL Photonics, Heidelberg, Germany). The protocol for recording images documenting filipin staining consisted of exciting the fluorophore with a Polychrome IV monochronometer at 380 nm with 1000-ms exposure time, utilizing a Fura filter cube (Chroma). Images were merged as RGB files using the TILLvisION software. The images were exported from the microscope software and processed with Adobe Pho- Statistical Analysis-The Resolver error model was used to build expression intensities. To determine significant differentially expressed genes within the groups, the TIGR MeV hierarchical cluster software (gift from the Canadian Bioinformatics workshop) was used for both genes and subjects. A p value of Ͻ0.05 was considered significant. The similarity metrics were based on Pearson correlation.
For TaqMan analysis, copy numbers of a given mRNA detected in each sample were calculated after normalization to the average log of the expression of three housekeeping genes (cyclophilin, glyceraldehyde-3-phosphate dehydrogenase, and B-actin). A one-way analysis of variance test, followed by Dunnett's test for multiple comparisons, was run separately for each gene to determine significant differences between groups.

RESULTS
Cathepsin D Gene Expression Is Decreased in Monocyte-derived Macrophages of Low HDL Subjects-Subjects were divided into four age-and sex-matched groups based on the presence or absence of low HDL-C, cholesterol efflux defect, and functional mutations in ABCA1 as follows: group A, low HDL, low efflux, no ABCA1 mutation; group B, low HDL, low efflux, ABCA1 mutation; group C, low HDL, normal efflux; group D, normal HDL, normal efflux. The control subjects (group D) had normal plasma levels of triglycerides and LDL cholesterol, similar to low HDL subjects and, by definition, higher HDL-C and hence higher total cholesterol concentrations. The three low HDL groups demonstrated very similar lipid values, and all groups were closely matched for age, sex, and body mass index (data not shown). 4 MDM from low HDL and control subjects were loaded with AcLDL; mRNA was isolated from the cells of each subject and analyzed using duplicate Affymetrix U133A oligonucleotide arrays (see "Experimental Procedures"). Comparative gene expression analysis revealed the differential regulation of several pathways reported elsewhere. 4 In particular, a cluster of cathepsin genes, which are major aspartic proteases of endolysosomes, was found to be down-regulated in MDM of low HDL-C groups compared with control ( Fig. 1A; red represents higher relative expression, and green represents lower relative expression for each gene comparing all of the subjects for that gene). CTSD was significantly decreased in efflux-defective subjects compared with control (p Ͻ 0.01) (Fig. 1B). These differences were confirmed by real time PCR (group A, p Ͻ 0.01; group B, p Ͻ 0.016), and both microarray results and real time PCR values were also recapitulated in a second cohort for group A (RT-PCR series 2, p Ͻ 0.05) but not for group B, for which no additional subject was available. The decrease in CTSD observed in the low HDL group with normal efflux (group C) was not significant in either microarray series 1 (Fig.  1, A and B) or series 2 or in RT-PCR assays. Similarly, ABCA1 expression was not significantly changed in groups A and C (data not shown). These observations led to the hypothesis that reduced expression of CTSD may play a role in plasma HDL levels.
CTSD Activity Modulates ABCA1-mediated Lipid Efflux-CTSD, an aspartic protease involved in the proteolytic degradation of proteins in endosome-lysosome compartments, binds ceramide with high affinity (37). The interaction of ceramide with CTSD results in autocatalytic proteolysis of the 52-kDa prepro-CTSD to form the enzymatically active 48/32-kDa isoforms of CTSD (37). Pepstatin is also a potent inhibitor of CTSD, which binds CTSD and inhibits its autocatalytic activity and generates less active forms of CTSD (38) (supplemental Fig. 1).
To investigate the effect of CTSD modulators on lipid efflux, J774 macrophages were radioactively labeled by incubation with LDL (100 g/ml) equilibrated with [   treated with ceramide, an activator of CTSD (37), or pepstatin A, a specific high affinity inhibitor of CTSD (38), for 24 h in DMEM containing 0.2% BSA, prior to the efflux assay. Ceramide increased the apoA-I-mediated efflux of cholesterol ( Fig. 2A) and phospholipid (Fig. 2B) in a dose-dependent manner by more than 80%. In contrast, inhibition of CTSD by pepstatin A resulted in a dose-dependent decrease in the efflux of both cholesterol (Fig. 2C) and phospholipid (Fig. 2D). Treatment with pepstatin A also impaired cholesteryl ester mobilization (Fig. 2E) and slightly decreased cellular free cholesterol (Fig. 2F). These experiments clearly show that the modulation of CTSD activity affects apoA-I-mediated efflux of cholesterol and phospholipid.
We next investigated whether the impaired apoA-I-mediated efflux by pepstatin A was associated with decreased ABCA1 protein. These studies demonstrated that treatment with pepstatin A (10 g/ml) significantly reduced ABCA1 protein in J774 macrophages (Fig. 3A) and to an even greater extent in CHO cells (Fig. 3B), where it was also accompanied by decreased cholesterol efflux (Fig. 3C). Concurrently, cell surface ABCA1 was also reduced upon treatment with pepstatin A (data not shown). These results are consistent with the hypothesis that suppression of apoA-I-mediated lipid efflux by pepstatin A is the result of decreased cellular ABCA1 protein. J774 macrophages were treated with or without a stimulator of cholesterol efflux, cAMP, for 24 h in the presence or absence of pepstatin A, and cholesterol and phospholipid effluxes were determined. In both conditions, pepstatin A significantly decreased apoA-I-dependent efflux of cholesterol (Fig. 4A) and phospholipid (Fig. 4B) in J774 cells. Pepstatin A similarly reduced cholesterol efflux from bone marrow-derived-macrophages (Fig. 4C) and in LDL or Ac-LDL treated peritoneal macrophages (Fig. 4D). These results indicate that CTSD promotes lipid efflux through the ABCA1-dependent pathway in primary macrophages and in immortalized J774 cells. Determination of cholesterol efflux and ABCA1 protein level in J774 cells at three doses of pepstatin demonstrated a highly significant correlation of the two parameters (data not shown).

Modulation of the Expression Level of CTSD Regulates ApoA-Imediated Lipid
Efflux-For a more direct assay of the effect of CTSD functional deficiency on apoA-Imediated lipid efflux, J774 cells were treated with siRNA CTSD or dsRNA CTSD for 36 h. The addition of CTSD-specific siRNA significantly reduced cholesterol efflux in a dose-dependent manner (Fig. 5A) and phospholipid efflux (Fig. 5B). At 100 nM, CTSD siRNA decreased cholesterol efflux by ϳ80% and phospholipid efflux by about 50%. In contrast, the scrambled dsRNA had no effect on these parameters.
Western blot analysis of CTSD and ABCA1 after siRNA treatment demonstrated decreased expression of CTSD with siRNA but not dsRNA (Fig. 5D) and parallel changes in ABCA1 protein mass (Fig. 5C). These results further support the notion that CTSD regulates ABCA1 levels and cholesterol efflux.
To explore the potential role of CTSD overexpression on lipid efflux and ABCA1 protein, CHO or J774 cells were transfected with a human full-length CTSD plasmid for 24 h, lipid efflux was assayed, and ABCA1 protein level was determined. Consistent with the siRNA results, transfection of human CTSD in CHO cells increased apoA-I-mediated lipid efflux (Fig. 6, A and B) and the cellular level of ABCA1 protein (Fig.  6C). Having shown that CTSD increases both ABCA1 protein and cholesterol efflux, we next determined whether CTSD exerted its effect directly on ABCA1 gene transcription. Macrophages were treated with either pepstatin A or by overexpression of human CTSD for 24 h, and ABCA1 mRNA levels were measured by RT-PCR. As expected, ABCA1 expression was significantly increased (Ͼ50%) by CTSD overexpression and was significantly decreased (Ͻ50%) by pepstatin treatment (Fig. 6D). Overall, these findings demonstrate that CTSD induces cholesterol efflux at least in part by regulating ABCA1 transcription. Because a small proportion of CTSD is secreted (25), we evaluated the possibility that overexpressed CTSD present in the medium could degrade the acceptors for cholesterol efflux. After efflux assay, medium apoA-I was immunoprecipitated and analyzed by Western blot, but there was no evidence of apoA-I fragmentation (data not shown). This may reflect the fact that CTSD is mostly active at low pH.
CTSD Is Involved in the Maturation of Saposins-Prosaposin (pro-SAP) is a 73-kDa precursor of four protein cofactors, saposins A, B, C, and D, which are required for the enzymatic hydrolysis of glycosphingolipids by specific lysosomal hydrolases (33,34). The proteolytic degradation of pro-SAP and the maturation to different saposins is dependent on CTSD (39). To examine whether CTSD regulates the proteolysis of pro-SAP in our model, we transiently overexpressed CTSD in CHO cells and quantified the expression of CTSD, pro-SAP, and the control protein HSP60 by Western blotting. As shown in Fig. 7, prosaposin was hydrolyzed into proteolytic products of ϳ10 kDa upon overexpression of human CTSD. Interestingly, we also found that overexpression of CTSD increased the level of both pro-SAP and saposins, and these increases were proportional to the level of CTSD (Fig. 7, top). These data verify that CTSD functions in the proteolytic processing of prosaposin. Interestingly, prosaposin mRNA expression was found to be decreased in the microarrays for series 1 and 2 for low HDL subjects compared with control subjects (p Ͻ 0.001 and p Ͻ 0.03, respectively).

Inhibition of CTSD Causes the Intracellular
Accumulation of Cholesterol and Sphingolipids-Since saposins are sphingolipid-binding proteins involved in the degradation of glycosphingolipids, we hypothesized that dysregulation of the expression of pro-SAP or its co-activator CTSD could impair glycosphingolipid (GSL) metabolism and cholesterol traffic. Indeed, the cellular transport and storage of cholesterol is linked to GSL metabolism (40 -43), and the accumulation of GSL in human macrophage foam cells and fibroblasts causes the trapping of cholesterol in endosome/lysosome compartments (41,43). Thus, in CHO cells, we assessed the effect of pepstatin treatment on the cellular localization of GSL using BODIPY-lactosylceramide and free cholesterol accumulation using filipin staining. We demonstrated perinuclear fluorescence of BODIPY-lactosylceramide (BODIPY Lac-cer) in control cells (Fig. 8A), whereas, in cells treated with pepstatin A, there was predominantly bright punctate endosome/lysosome fluorescence (Fig. 8B) similar to the cellular phenotype of Niemann-Pick type C disease (NPC) (44). Extensive studies have also shown the accumulation of GSL and cholesterol in late endosomes/lysosomes in NPC cells (45). To test whether the addition of exogenous ceramide would also enhance cholesterol efflux in NPC1-deficient cells, we isolated bone marrow-derived macrophages from control and NPC1-deficient mice and measured apoA-I-mediated cholesterol efflux. The addition of ceramide (20 M) or cAMP increased cholesterol efflux in control cells (2-fold) but not in the NPC1-deficient cells (Fig.  8C). As expected, pepstatin also tended to cause enlarged punctate fluorescence of filipin-stained free cholesterol in endosome/lysosome-like structures, whereas in control cells cholesterol distribution was diffuse throughout the cell (data not shown). Taken together, these results demonstrate that inactivation of CTSD function results in the accumulation of GSL and free cholesterol in late endosomes/lysosomes.
Inhibition of CTSD Causes the Mistargeting of ABCA1 Protein to Late Endosomes/Lysosomes-ABCA1 is normally localized at the cell surface and in the late endosomal com-  DECEMBER 29, 2006 • VOLUME 281 • NUMBER 52 partment (46). To test whether inhibition of CTSD by pepstatin affected ABCA1 localization, we followed the distribution of ABCA1 in CHO cells by confocal microscopy using GFP-ABCA1 and Lysotracker red as a lysosomal marker. In these experiments, cycloheximide treatment was omitted to avoid interference with the degradation of lysosomal enzymes. Under these conditions, we observed no co-localization of ABCA1 with Lysotracker red (Fig. 9A) in control cells, but in cells treated with pepstatin, ABCA1 was found to colocalize extensively with lysosomes, as indicated by the yellow punctate fluorescence (Fig. 9B). These results show that inhibition of CTSD activity causes trapping of ABCA1 in a lysosomal compartment, impairing its trafficking to the plasma membrane.

DISCUSSION
These studies were designed to identify and evaluate novel candidate genes relevant to HDL metabolism and the etiology of low HDL syndromes. By focusing on genes differentially expressed in a group of subjects with low HDL, we identified a cluster of cathepsin genes. Notably, mRNA expression of CTSD, the major aspartic protease of endolysosomes (25), was decreased in the low HDL groups as compared with control subjects and significantly in groups with low efflux (p Ͻ 0.01) (Fig. 1). These results were confirmed in a second cohort that included a group with low HDL and efflux defect but no mutation in ABCA1, whereas the group with low HDL and normal efflux did not present with decreased CTSD. Although the decreased CTSD could not be verified in a second group with low HDL and efflux with ABCA1 mutations (for lack of subjects), the initial observation (Fig. 1B) suggests the possibility of an independent decrease in CTSD that compounds the ABCA1 defect. Alternatively, the decreased ABCA1 activity may decrease CTSD through a feedback loop.
Based on these observations, we hypothesized that CTSD activity might affect cellular cholesterol transport or efflux. CTSD binding of ceramide affects its autocatalytic activity (37), and in cells with acid sphingomyelinase deficiency, low endogenous ceramide (the product of the digestion of sphingomyelin) is associated with decreased CTSD activity (37). We showed that exogenous ceramide increases cholesterol and phospholipid efflux by increasing the autocatalytic activity of CTSD (supplemental Fig. 1). In agreement with our results, Witting et al. (47) have shown that the addition of ceramide increases cholesterol efflux and cell surface ABCA1. Alternatively, ceramide may also affect the cholesterol efflux through the feedback activation of acid sphingomyelinase (48).
CTSD has been shown to bind and stimulate the proteolytic processing of proSAP (39). The proSAP knock-out mice (49) and tissues from individuals with proSAP deficiency (50,51) are characterized by GSL accumulation (40 -43). Recently, our laboratory showed that the inhibition of prosaposin delivery to late endosomes by LRP1 and its adaptor, GULP, causes the accumulation of GSL and free cholesterol in late endosomes, resulting in down-regulation of ABCA1 and decreased cholesterol efflux (52). Consistent with these observations, we demonstrate that the inhibition of CTSD is also accompanied by the accumula-  tion of both BODIPY-lactosylceramide and free cholesterol in structures resembling endosome/lysosome compartments reminiscent of the cellular phenotype of NPC disease (44).
The identification of CTSD as an activator of pro-SAP has provided important insight into the mechanism by which CTSD modulates the metabolism of cellular cholesterol and GSL. Pro-SAP is the precursor of four protein cofactors, saposins A, B, C, and D, that are required for the enzymatic hydrolysis of sphingolipids by specific lysosomal hydrolases. Saposins have important functions in the degradation of ceramide (saposin D), glucocerebroside (saposin C), sulfatide (saposin B), and glycosylceramide (saposin C) (53). Recently, saposin B was also shown to bind and transfer phospholipids from one membrane to another (54), a process that can occur in lysosome/late endosome compartment and is specific for either phosphatidylcholine or phosphatidylinositol. Here, we show that overexpression of CTSD increases cellular protein levels of both pro-SAP and saposins, which control sphingolipid transport and the cholesterol efflux pathway. Sphingolipidoses are lysosomal storage diseases. One of the most common, Gaucher's disease, is caused by deficiency of glucocerebrosidase, a lysosomal acid ␤-glucosidase, which results in the accumulation of glucosylceramide in lysosomes. Notably, affected subjects have decreased plasma concentrations of HDL cholesterol and apoA-I (55,56). Niemann-Pick, type A or B, is caused by altered function of acid sphingomyelinase, which metabolizes the sphingomyelin to ceramide. The saposins A, C, and D are responsible for the hydrolytic activation of this pathway (57,58). Individuals with this disease also exhibit low HDL-cholesterol levels (59), as do patients with NPC disease (60). Furthermore, it has been reported that macrophages from acid sphingomyelinase knock-out mice accumulate sphingomyelin and exhibit defective cholesterol trafficking and efflux (61). Fibroblasts treated with lactosylceramide or with an inhibitor of glucocerebrosidase also accumulate glycosphingolipids, which results in inhibition of cholesterol efflux by the apoA-I/ ABCA1 pathway (62). Thus, disruption of one of the pathways FIGURE 9. Pepstatin causes mistargeting of ABCA1 protein that colocalizes with late endosomes/lysosomes. CHO cells were grown in 10% FBS and transfected with ABCA1-GFP for 24 h in the absence or presence of 10 g/ml pepstatin (panels A and B, respectively). The cells were then incubated with Lysotracker red, a lysosomal marker, for 10 min at room temperature and visualized by confocal microscopy. Confocal images of lysosomes are displayed in red, and GFP-ABCA1 is shown in green. In the presence of pepstatin ABCA1 colocalizes with Lysotracker in a lysosomal compartment. of sphingolipid metabolism can cause GSL and free cholesterol accumulation and alter cellular lipid homeostasis and HDL concentrations (Fig. 10). In this study, the low HDL subjects with efflux defects have low CTSD expression, suggesting a correlation between CTSD expression and HDL levels. This low HDL phenotype could also be additive with decreased expression of prosaposin (50). ABCA1 expression levels are not significantly decreased in group A (data not shown), suggesting that decreased CTSD may also affect efflux by dysregulation of ABCA1 trafficking rather than expression.
The expression and regulation of cathepsins are complex. They can act as transcriptional activators (63), and there can be cross-talks between cathepsins, where for example cathepsin L is involved in CTSD processing and regulation (64). The CTSD gene is controlled by a mixed promoter that contains features of both housekeeping and regulated genes (65)(66)(67)(68). It is regulated by estrogens (66,69) and by retinoids (65,70). Here, a decrease in CTSD results in decreases in saposins, which translate in glycosphingolipid accumulation in late endosomes that in turn blocks cholesterol transport, preventing regulation of cellular cholesterol and decreasing the efflux pathway. Thus, the regulation of CTSD in monocyte-macrophages by retinoic acid (65) is compatible with the general contribution of liver X receptors and retinoid X receptors to the control of cholesterol homeostasis and inflammation (71).
In conclusion, these studies show for the first time that CTSD regulates intracellular transport of phospholipid and cholesterol and ABCA1-mediated lipid efflux. The addition of pepstatin A, an inhibitor of CTSD, impaired apoA-I-mediated efflux in CHO cells and in macrophage cell lines. Specific siRNA oligonucleotides had a similar effect, causing a marked reduction in cholesterol and phospholipid efflux. Conversely, transfection of murine macrophages or CHO cells with a CTSD cDNA expression plasmid led to a moderate but significant increase in apoA-I-mediated lipid efflux. In addition, both mRNA and protein expression of ABCA1 was induced by CTSD and inhibited by pepstatin A. Taken together, these findings suggest that decreased CTSD expression can contribute to low HDL. These findings are relevant to our understanding of the regulation by lysosomal enzymes of intracellular cholesterol trafficking and efflux.